Reconstruction comparison and a composite segmentation method for electron tomography

Electron Tomography (ET) reconstruction involves solving an inverse problem to generate a volumetric map from multiple projection images (tilt-series) acquired from an Electron Microscope (EM). Due to limitations in the acquisition process no unique solution exists for the inverse problem. Hence the reconstructed volumes are different depending on the choice of reconstruction method used. Here, we analyze and compare reconstructions from four different techniques. Secondly, a novel technique for composite segmentation is presented where reconstructed volumes from four different methods and their segmented maps are combined. This method annotates the segmented output with a confidence level based on a voting scheme. This enhances structure visualization and provides comprehensive structure information with error and uncertainty annotations at a voxel level.